if rolling-update is forced to delete old pod, some responses about old pod will be interrupted. Example: Rc (myapp) define replicas 2 and contain 2 pods(myapp-mtlz8, myapp-an78c). rolling-update rc (myapp). When create a new pod(myapp-ed4988fb0b53ed961037a026068d1a3d-i8wvt) and start to delete a old pod(myapp-mtlz8). If old pod is being processed some requests, but the old pod is deleted. So some request don't complete about old pod.

Whether or not: 1. service don't proxy to old pod 2. Waiting to old pod completed all request 3. delete old pod